KR102514798B1 - Computing system for quantitative pricing-based task offloading of iot terminal considering latency in mobile edge computing environment, and method thereof - Google Patents

Abstract

Various embodiments relate to a computer system and method for offloading a task of an IoT terminal in consideration of latency in a quantitative cloud-based edge computing environment, detecting an electronic device to be offloaded, and detecting a computer system for the electronic device. Proposes a price bidding function to set a unit price per second for CPU usage, receives CPU processing time of the computer system for use in the electronic device from the electronic device, and processes tasks offloaded from the electronic device during the CPU processing time. can be configured to

Description

Computer system and method for task offloading of IoT terminal considering latency in quantitative cloud-based edge computing environment

Various embodiments relate to a computer system and method for task offloading of an IoT terminal considering latency in a quantitative cloud-based edge computing environment.

Recently, artificial intelligence (AI) technology for analyzing Internet of Things (IoT) data is developing rapidly. Accordingly, numerous mobile devices such as smartphones, tablet PCs, and wearable devices are appearing and gaining popularity. Advances in technology and the popularity of mobile devices have increased the demand for computationally intensive IoT applications such as 3D games and virtual reality (VR). You can think of execution delays, also known as latency, as long as the task completion time required to run the application. Computationally intensive applications need to handle many tasks in a short amount of time, so these applications are sensitive to execution latency. However, mobile devices have limited computing power capable of providing a satisfactory quality of experience (QoE).

In this regard, cloud computing was introduced to solve the problem of limited functionality of mobile devices by distributing computer workloads to centralized servers. Devices with relatively little computational capacity send part of their computational workload to the server, which processes it and returns results. The device then receives and uses the results instead of processing the workload itself. By using cloud computing technology, mobile devices are considered capable of meeting the execution delay requirements of computationally intensive applications. However, distributing the computing workload, called offloading, introduces additional latency due to communication. Therefore, existing cloud computing has limitations in reducing latency because the centralized server is physically far from the user.

In order to reduce this latency caused by communication, Mobile Edge Computing (MEC) has been introduced which sends intensive task loads to MEC servers located at the edge of the cellular network. Since these MEC servers are located closer to users than existing centralized servers, mobile edge computing can further reduce waiting time due to communication. A study on mobile edge computing presented an optimization problem to achieve optimal resource allocation and solved the problem to maximize energy efficiency or minimize execution delay. Resources can be considered as computational capacity, transmission strength, network bandwidth, etc. However, most of these studies focus only on improving performance throughout the system, causing many problems in the fairness of IoT devices. In this regard, many studies have been conducted to solve this problem. Still, this effort to improve fairness doesn't adequately portray some users who want to use more edge servers, even at higher prices. In this respect, it can be a good choice to allocate the amount of resources according to the user's desire through a pricing mechanism. From this point of view, recently, several studies considering the price of the MEC environment have been conducted.

However, most existing tasks have focused on the amount of data offloaded in pricing. Thus, the user's payment is proportional to the amount of offloaded data based on the existing pricing model. Therefore, if a user offloads the same amount of data, he or she will pay the same amount regardless of the server's computational resource usage. Therefore, users will prefer to use a lot of computational resources because the payment is the same no matter how much computational resources are used. However, from the server's point of view, since the server's computational resources are limited, each user's computational resource usage must be taken into account when determining the price. If users pay in proportion to the amount of offloaded data as well as the amount of server computational resources, the server does not have to worry about overload due to excessive use of resources and can operate more efficiently by providing limited resources to more users. be able to Therefore, the pricing method should consider the use of computational resources and the amount of offloaded data.

Various embodiments model a quantitative pricing plan in which the price is determined according to the capacity and data used according to the movement of the current cloud industry, and how to determine the optimal offloading of users (resource usage, data distribution) in an environment in which the user has the right to determine resource usage. suggests According to various embodiments, users can determine an optimal server usage capacity and an amount of offloading data capable of processing data at a low cost while meeting a latency condition of an application they are running.

Various embodiments provide a computer system and method for task offloading of an IoT terminal considering latency in a quantitative cloud-based edge computing environment.

A method of a computer system according to various embodiments includes detecting an electronic device to be offloaded, proposing a price bidding function to set a unit price per second for CPU usage of the computer system for the electronic device; Receiving CPU processing time of the computer system for use in the electronic device from the electronic device, and processing a task offloaded from the electronic device during the CPU processing time.

A computer system according to various embodiments includes a memory and a processor connected to the memory and configured to execute at least one command stored in the memory, wherein the processor detects an electronic device to be offloaded; Propose a price bidding function for setting a unit price per second for CPU usage of the computer system for the electronic device, receive a CPU processing time of the computer system for use in the electronic device from the electronic device, and During processing time, it may be configured to process tasks offloaded from the electronic device.

According to various embodiments, a non-transitory computer-readable storage medium may include a step of detecting an electronic device to be offloaded, a price bidding function for setting a unit price per second for CPU usage of the computer system for the electronic device. Proposing a, receiving from the electronic device CPU processing time of the computer system for use in the electronic device, and processing a task offloaded from the electronic device during the CPU processing time. can store one or more programs for

According to various embodiments, according to the movement of quantitative pricing in the current cloud industry, users can increase Quality of Service (QOS) by using an offloading service at low cost, and accordingly, a large-scale user influx into the cloud service industry. can be expected Accordingly, since users can run high-performance applications with mobile devices that have limitations in computing performance, the demand for high-performance applications increases, which can be expected to activate the game or artificial intelligence-based mobile application market.

1 is a diagram illustrating a system model according to various embodiments.
2 is a diagram illustrating an overall process in a system model according to various embodiments.
3 and 4 are diagrams for explaining a price bidding function according to various embodiments.
5 is a diagram illustrating a neural network that determines a price bidding function in accordance with various embodiments.
6 is a diagram illustrating an algorithm for determining a price bidding function in accordance with various embodiments.
7 is a diagram illustrating a computer system and an electronic device of a system model according to various embodiments.
8 is a diagram illustrating a method of a computer system in accordance with various embodiments.
9 is a diagram illustrating a method of an electronic device according to various embodiments.

Hereinafter, various embodiments of this document will be described with reference to the accompanying drawings.

1. System model

1 is a diagram illustrating a system model according to various embodiments.

Referring to Figure 1, a system model according to various embodiments considers a situation where one server is already used by some users and new users are offloaded. The user's goal is to offload while paying to the server to minimize execution delay and meet the application's execution delay requirements. Here, it is assumed that the application can be partially processed. One of the applications is 3D rendering, which is a 3D computer graphics process that transforms a 3D model into a 2D image on a computer. This 3D rendering is done on an object-by-object basis and can be partially processed for each object. It is assumed that data transmission between the server and the user's mobile device follows the LTE standard.

2 is a diagram illustrating an overall process in a system model according to various embodiments.

Referring to FIG. 2 , a user has information about a local CPU capacity (F _l ), an amount of data to be processed (d), an execution delay requirement (treq), and a workload (h) according to an application. Here, workload is defined as the number of CPU cycles required to process one bit of data. When a new user wants to offload, the user reports the information mentioned above. Then, the server considers this information to find an appropriate price and proposes a determined price to the user. A proposed price bidding function is used in this process, which will be described later. Finally, the user makes an offloading decision on server offloading and CPU usage to minimize pay-as-you-go execution delays. The method of determining the price bid function of the server is to determine an appropriate price bid function to maximize profits. Regarding the price bidding function, it will be described later in more detail with reference to Figs. 3, 4, 5 and 6. The user's optimal offload decision method is to determine the server's offload and CPU usage to minimize execution delay considering payment. In addition, the configuration and operation of the server and the user in the system model shown in FIGS. 1 and 2 will be described in more detail later.

3 and 4 are diagrams for explaining a price bidding function according to various embodiments. 5 is a diagram illustrating a neural network that determines a price bidding function in accordance with various embodiments. 6 is a diagram illustrating an algorithm for determining a price bidding function in accordance with various embodiments.

A. Payment model

Various embodiments propose a differential pricing scheme to solve the problems of existing schemes. According to the existing task pricing model, payment is proportional to the size of the task to be processed. These models only consider the number of CPU cycles required to process a task, regardless of the usage of CPU capacity, that is, the number of CPU cycles per second. Therefore, all users want to use a lot of CPU capacity, which leads to inefficiency because the CPU capacity of the server is limited. A pricing method according to various embodiments determines a price based on a CPU capacity used by a user. Thus, the user's payment is based on both the server CPU capacity used by the user and the size of the tasks offloaded.

Here, we assume that the user decides how much CPU capacity to use on the server and uses this capacity until the offloaded data is completely processed. In this case, the payment is the product of the processing time of the server and the unit price per second.

Various embodiments propose a price bidding function that takes CPU usage into account in the price used to set the unit price per second. To solve the problem mentioned above, the price bidding function should make the unit price higher because the user uses more CPU capacity of the server. Let F _T represent the total CPU capacity of the server and F _a represent the CPU capacity of the server used by the user. F _a /F _T means the user's CPU usage rate. As shown in FIG. 3 , in the pricing method according to various embodiments, the unit price per second is set as the product of the CPU usage amount and the value of the price bidding function according to the input of the usage amount, as shown in [Equation 1] below.

where ω is the unit price per second and f is the price bidding function.

As shown in Fig. 4, since the user uses the server during the processing time, the product of the unit price and the processing time is the total payment. Suppose the user offloads l bits out of d bits. The number of CPU cycles required to process one bit depends on the application. Let this value be denoted by h. Then, hl represents the total number of CPU cycles required to process the offloaded data. The processing time indicated by t _p can be expressed as in [Equation 2] below.

The total payment is the product of unit price and processing time. Unit prices are indicated by gray squares in FIGS. 3 and 4 . Therefore, the payment to be paid by the user can be expressed as the following [Equation 3] based on [Equation 1] and [Equation 2], which is equal to the volume of the hexahedron shown in FIG. 4 . When the price bidding function is linear, the payment W can be expressed as in [Equation 4] below.

where W is the total payout.

Here, the parameter b must be greater than 0 because it must always have a positive value, and the parameter a must be greater than 0 because the user uses more CPU capacity of the server and therefore must be more expensive.

Lemma 1. Payout W is an increasing function for l and F _a, respectively.

proof. h is the required number of CPU cycles for one bit of data, and F _T remains the CPU capacity of the server. They must be positive. Parameters a and b must be positive by definition. The partial derivative of the payout with respect to l and F _a is

Thus, the payout W is an increasing function for l and F _a respectively.

Based on a payment model with a linear price bidding function, payment increases with user usage of server CPU capacity as well as amount of offloaded data. For example, if two users are offloading the same amount of data, 200kbytes, and using the same application with a workload of 1000, their pay will vary based on CPU usage. Assume that one user uses 1Х10 ⁹ CPU cycles per second, another user uses 2Х10 ⁹ CPU cycles per second, and the total CPU capacity of the server is 100Х10 ⁹ cycles per second. Their payouts will be 0.248 and 0.256. where a and b are 100 and 30. Therefore, users who use more of the server's CPU capacity will have to pay more.

The price bidding function can be any incremental function. If the server has a discount policy, the log function can be a price bid function. The exponential function could be the price bidding function of the server with the most user support policy. This paper deals with linear functions that can include the mean of these policies.

B. Stackelberg game model

In this section, servers and users are formulated as a single leader-many-follower Stackelberg game model, in which the server acts as the leader and the user acts as the majority predictor. To do this, we first define the user's cost function and the server's utility function. Each user minimizes its own cost function, and the server maximizes its utility function.

When the server proposes a price bid function, the user can decide how much data to offload and how much of the server CPU capacity to use. Since the user's goal is to reduce the execution delay, the user cost must include the execution delay. Also, since users want to minimize execution delays for small amounts, we need to include an additional term to refer to payment. The more the execution delay the user wants, the more they have to pay according to the proposed payment model. Therefore, users must consider the trade-off between delay in execution and payment. In other words, user costs must consider both execution delays and payments.

First, we will model the execution delay and define the user cost function using the payment model presented in Section II-A.

The user offloads part of the data and processes the rest of the data locally at the same time. Thus, the execution delay can be expressed as the long time between the time spent offloading and the local processing time.

The execution delay is denoted by t _e and t _off , t _loc as offload and local processing time, respectively. Then, the execution delay can be expressed as:

The time required for offloading can be set as the sum of uplink data transmission time (t _up ), server processing time (t _p ), and downlink data transmission time (t _down ).

Since the communication between the MEC server and the mobile device is a one-hop communication, it can be considered that the network capacity does not have a significant effect on the performance. Therefore, we do not consider network capacity, but rather assume that the bandwidth and data transmission capacity are the same for all users. Therefore, the data rate, which is the number of bits that can be transmitted per second, is a constant. Uplink and downlink data rates are denoted as r _up and r _down respectively.

The uplink data transmission time can be expressed as l/r _up . The data size after processing will be different from the size before processing, and this ratio is assumed to be r _io . Then, the downlink data transmission time can be expressed as r _io l/r _down . By using this, based on [Equation 6] above, the time required for offloading can be expressed as shown in [Equation 7] below.

Since the user has to process the dl bit locally, (dl)h is the number of CPU cycles required for local processing. The local processing time (t _loc ) can be expressed as in [Equation 8].

where F _l is the user's local CPU capacity. The execution delay can be set as follows based on [Equation 5] to [Equation 8] above.

Definition 1 (user's execution delay). The user's execution delay is defined as in [Equation 9] below.

Since the user's goal is to reduce execution delay with microtransactions, we can consider the trade-off between them by setting the user cost function to be the weighted sum of execution delay and payment. The decision variables for the user's optimal offloading decision are the amount of data to be offloaded (l) and the CPU capacity of the server to be used (F _a ). In this regard, the user cost function is set as follows.

Definition 2 (user cost function). The user cost function is defined as in [Equation 10] below.

Here, β is a weight coefficient that reflects the user's willingness to pay (β > 0).

A small β means that users think that delay in execution is more important than payment. That means users are more willing to pay. By minimizing this user cost function, the user can make optical offloading decisions that minimize the execution delay taking into account payment.

In the system model according to various embodiments, the server obtains revenue by providing an offloading service and seeks to maximize revenue. Therefore, the server must determine an appropriate price bidding function to maximize profits. At this time, the price bidding function is regarded as a linear function expressed as f(x) = ax + b, and a and b are decision variables of the server.

와

가 사용자의 최적 오프로딩 결정이라 하자. 그런 다음 서버 유틸리티를 다음과 같이 정의한다.Once the server proposes a price bid function, the user makes an offloading decision based on it. Here, it is assumed that the user always optimizes offloading decisions, and the server knows the user's optimal strategy.

and

Let be the user's optimal offloading decision. Then define the server utility as:

Definition 3 (Server Utility Functions). Server utility is defined as revenue, which is expressed as in [Equation 11] below.

Server utility is defined as its return in Definition 3, and the purpose of a server is to maximize utility.

Finally, we formulate a single leader-many-follower Stackelberg game based on the user's cost function and the server's utility function. Servers are leaders, IoT users are followers.

2. Equilibrium Assay

In the following, the goal balancing strategy of the user and the server is derived based on the Stackelberg game formulated above.

A. Your optimal strategy

In this section, we propose an optimal offloading decision-making system for users. Based on the server pricing strategy, minimizing costs is the user's optimal strategy. When the server proposes a price bid function, the user can decide how much data to offload and how much of the server's CPU capacity to use. The user's objective is to minimize its costs, consisting of delays in execution and payments. Therefore, with F _a and l as decision variables, the optimization problem can be formulated as follows.

Problem 1 (optimal offloading decision problem)

Here, F _r is the remaining CPU capacity of the server when the user wants to offload, t _req represents the user's execution delay requirement, and t _e is the execution delay expressed in [Equation 9] above. The first constraint (st) means that the offloaded data size cannot be less than 0 or greater than the total data size. The second constraint means that the CPU usage of a user on the server cannot be less than 0 and cannot be greater than the remaining CPU capacity of the server at that time. The third constraint means that the execution delay must be less than or equal to the user's requirement.

이 사용자가 서버에서 연산 용량 F_a를 사용할 때 오프로드에 소요되는 시간과 로컬 처리 시간을 동일하게 하는 데이터 크기를 나타낸다 하자. F_a의 함수로 표현할 수 있으며, 우리는 다음과 같이 정의한다.

Let us denote a data size that equates the time required for offloading with the local processing time when this user uses the computing capacity F _a on the server. It can be expressed as a function of F _a , and we define it as:

(F_a)으로 표시한다. Definition 4 (balanced data size to offload). As shown in [Equation 12], the equilibrium data size is defined as the data size for which the offload time and the local processing time are equal,

(F _a ).

이다.where A is

am.

(F_a) 미만일 경우, 로컬 처리 시간이 오프로드에 소요된 시간보다 길다. 로컬 처리 시간은 l가

(F_a)보다 클 때 오프로드에 소요된 시간보다 짧다.l is

If less than (F _a ), the local processing time is longer than the time taken for offloading. Local processing time is l

(F _a ) is shorter than the time spent off-road.

를 사용자 비용 함수로 한다.To this end, when the user uses the computing capacity F _a 'in the server

is the user cost function.

로 사용할 때 사용자 비용 함수를 하기 [수학식 13]과 같이 정의한다. Definition 5 . The user sets the computational capacity Fa' of the server

When used as , the user cost function is defined as in [Equation 13].

The second term of the user cost function, the user payment, increases based on lemma 1. Depending on the slope of the pay term with respect to l, the cost function may have a different shape.

보다 크면 사용자 비용 함수

가 증가 함수가 될 것이다. Lemma 2. If the payout slope for l is

greater than the user cost function

will be an increasing function.

proof. Based on the data size (l) to be offloaded, the user's execution delay can be expressed as:

The partial derivative of the user cost function with respect to l is

인 l에 대한 지불 기울기가

보다 크다고 가정하자. 그러면,

is the payout slope for l

Let's assume it's bigger than then,

이면, 비용 함수

은 증가 함수이다.

If , the cost function

is an increasing function.

보다 작을 경우 사용자 비용 함수

는

에서

와 같이 최소값을 갖는다. Lemma 3. If the payout slope for l is

User cost function if less than

Is

at

has a minimum value as

의 편도함수는 보조정리 2의 증명에 제시되어 있다.

라 가정하자.proof. user cost function

The partial derivative of is given in the proof of Lemma 2.

Let's assume

값은 l가 ~l(F0a)까지 증가할수록 감소하고, l이

이상 증가할수록 증가한다. 따라서 사용자 비용 함수는

에서 최소값을 갖는다.user cost function

The value decreases as l increases up to ~l (F0a), and when l

increases as the number increases. So the user cost function is

has a minimum value in

과 같을 경우, 사용자 비용 함수

는 모든

에 대해 동일한 값을 가지며, 모든

에 대해 증가 함수가 될 것이다. Lemma 4. If the payout slope for l is

, the user cost function

is all

has the same value for all

will be an increasing function for

라 가정하자. 그러면,proof.

Let's assume then,

It is based on the partial derivative of the user cost function in the proof of Lemma 2.

에 대하여 동일한 값을 가지고, l에 대해 l이

이상 증가하면 증가한다.Therefore, the user cost function is

has the same value for , and l for l

Increases more than

과 같다면 서버에서 초당 Fa' 컴퓨팅 사이클을 사용하는 경우

을 보내는지 여부에 따라 사용자는 동일한 비용을 갖는다. 그러나 오프로딩 시 사용자는 에너지 소비의 혜택을 받을 수 있고 로컬 컴퓨팅 시간을 줄일 수 있어 다른 태스크를 처리할 수 있는 시간이 주어진다. 따라서 l에 대한 지불 기울기가

과 같을 때 사용자 비용이 같더라도 사용자가

비트의 데이터를 오프로드한다고 가정한다.Based on Lemma 4, the payout gradient for l is

If the server uses Fa' computing cycles per second, then

Depending on whether or not to send the user has the same cost. However, when offloading, users can benefit from energy consumption and reduce local computing time, giving them time to tackle other tasks. Therefore, the payout slope for l is

When , even if the user cost is the same, the user

Assume we offload bits of data.

과 같을 경우, 사용자는

전송 여부와 상관없이 사용자 비용이 동일하더라도 데이터의

비트의 데이터를 오프로드한다. Assumption 1. If the payout slope for l is

, the user

of data, even if the user costs the same regardless of whether it is transmitted or not.

Offload bits of data.

Then, theorem 1 can be derived from lemmas 2, 3, 4 and assumption 1.

이면, 오프로딩에서 혜택을 가질 수 있다. 그러지 않다면, 오프로딩은 사용자에게 혜택을 줄 수 없다. Theorem 1. Users

, you may benefit from off-roading. Otherwise, offloading cannot benefit users.

이면, 사용자 비용 함수는 l에 대하여 증가 함수이다. 즉, 오프로드된 데이터의 크기가 커질수록 사용자 비용이 증가하는 것이다. 따라서 사용자가 오프로드하지 않는 것이 최적이다.

이면 사용자는 보조정리 3에 근거한 오프로드의 혜택을 얻을 수 있다. 보조정리 4와 가정 1에 따르면 사용자 역시 오프로드의 혜택을 누릴 수 있다.proof. Based on the lemma 2,

, then the user cost function is an increasing function with respect to l. That is, as the size of the offloaded data increases, the user cost increases. Therefore, it is optimal that the user does not offload.

If this is the case, the user can benefit from offloading based on lemma 3. According to Lemma 4 and Assumption 1, users can also benefit from offloading.

We can now extend Fa' to Fa without loss of generality. Then, in the user cost function V(Fa, l), it is known that the user can receive an off-road benefit when the condition as shown in [Equation 15] below is satisfied.

This can be re-expressed as in [Equation 16] below.

만큼 오프로드하는 것이 최적 오프로드 결정이다. 그렇지 않으면 사용자는 오프로드의 혜택을 받을 수 없으므로 사용자의 최적 전략은 오프로드하지 않는 것이다.In the case of l, if the condition as in [Equation 16] is satisfied,

Offloading as much as possible is the optimal off-road decision. Otherwise, the user cannot benefit from offloading, so the user's optimal strategy is not to offload.

는 사용자가 서버에서 F_a 계산 용량을 사용할 때 오프로드할 수 있는 최적의 데이터 크기이다. 상기 [수학식 16]과 같은 조건이 충족되면 F_a에서도 최적의 오프로드 결정을 찾아야 한다. 따라서 앞으로는 사용자가

를 상기 [수학식 16]과 같은 조건으로 오프로드할 때 최적의 F_a를 찾는다.in this respect

is the optimal data size that can be offloaded when a user uses F _a computing capacity on the server. If the condition as in [Equation 16] above is satisfied, an optimal offload decision must be found in F _a as well. Therefore, in the future, users

When offloading is performed under the same conditions as in [Equation 16] above, the optimal F _a is found.

비트의 데이터를 오프로드하면 실행 지연은 로컬 처리 시간과 동일해진다. 문제 1의 세 번째 제약 조건은 하기 [수학식 18]과 같이 다시 쓸 수 있다.If the condition as in [Equation 16] above is satisfied, the user

By offloading a bit of data, the execution delay equals the local processing time. The third constraint of Problem 1 can be rewritten as in [Equation 18] below.

This is as shown in [Equation 19] below.

비트의 데이터를 오프로드할 때 상기 [수학식 14]에 근거하여 F_a에 대한 최적 오프로드 결정 문제를 다음과 같이 수정할 수 있다.After that, the condition as in [Equation 16] is satisfied, and the user

When bit data is offloaded, the optimal offload decision problem for F _a can be modified as follows based on [Equation 14] above.

Problem 2 (Fix Optimal Offload Determination Problem).

Here, the constraints for [Equation 20] are the same as the second and fourth constraints of Problem 1, [Equation 16] and [Equation 19] above.

을 오프로딩할 때 사용자 비용 함수로, 사용자가 서버에서 계산 용량 F_a를 사용할 때 오프로딩할 데이터 크기에 대한 최적의 오프로드 결정이다. 우리는 이 비용 함수를

으로 정의한다. In [Equation 20], the correction cost function is

is the optimal offload decision for the size of data to offload when the user uses the computational capacity F _a on the server. We use this cost function

Defined by

비트를

로 오프로딩할 때 사용자 비용 함수를 하기[수학식 21]과 같이 정의한다. Definition 6. Users of data

bit

When offloading to , the user cost function is defined as follows [Equation 21].

It is the same as [Equation 20] above.

은 Fa≥0에서 볼록함수로, 다음과 같이 설명한다.user cost function

is a convex function at Fa≥0, and is described as follows.

는 Fa≥0에서 볼록 함수이다. Lemma 5. User cost function

is a convex function at Fa≥0.

의 2차 편도 함수는 다음과 같이 표현된다.proof. user cost function

The second order partial derivative of is expressed as

는 항상 0보다 크거나 같기 때문에 비용함수의 2차 편도함수는 항상 F_a≥0이다. 따라서 비용함수

는 Fa≥0에서 볼록 함수이다.According to the second constraint of Problem 1 and the same conditions as [Equation 16] above,

is always greater than or equal to 0, so the second partial derivative of the cost function is always F _a ≥ 0. Therefore, the cost function

is a convex function at Fa≥0.

Therefore, this correction problem is a convex optimization problem with convex constraints, and the solution to minimize the correction cost function is described in Lelem 6.

는 Lemma 6. A solution that minimizes the modified cost function

Is

이다.where B is

am.

의 1차 편도 함수는 다음과 같이 표현된다.proof.

The first partial derivative of is expressed as

를 1차 편도 함수가 0과 동일하게 만드는 값을 나타낸다 하자. 그러면 이는 다음과 같이 표현된다.

Let denote the value that makes the first-order partial derivative equal to zero. Then this is expressed as:

는 항상 양의 값을 가진다. Fa가 Ca≥0에서

보다 작다면, 비용은 감소하고 그렇지 않으면 증가한다.

는 수정 비용 함수

를 최소화한다.Since B always has a negative value based on the second constraint of Problem 1 and the same condition as [Equation 16] above,

always has a positive value. When Fa is Ca≧0

less than , the cost decreases and otherwise increases.

is the modified cost function

to minimize

는 실현 가능한 범위를 벗어나면 최적의 해결책이 아닐 수 있다. F_min은 하한을 나타내며, F_max는 문제 1의 두 번째 제약 조건, 상기 [수학식 16], 상기 [수학식 19]에 근거하여 실현 가능한 F_a 범위의 상한을 나타내도록 한다.however

may not be an optimal solution if it is outside the feasible range. F _min represents the lower limit, and F _max represents the upper limit of the feasible F _a range based on the second constraint of Problem 1, [Equation 16] and [Equation 19] above.

는 정리 2에 의해 설명된다.Based on these feasible limits, the best feasible solution

is explained by Theorem 2.

는 상기 [수학식 17], [수학식 23], [수학식 24], 보조정리 5 및 보조정리 6에 근거하여 다음과 같다. Organize 2. A feasible optimal solution that minimizes the modified cost function

is as follows based on [Equation 17], [Equation 23], [Equation 24], lemma 5 and lemma 6 above.

이라는 조건이 충족되어야 한다. 그렇지 않다면 실현 가능한 해결책은 존재하지 않는다. 즉, 사용자의 최적 전략은 오프로드하지 않는 것이므로 실현 가능한 최적 솔루션은 (0, 0)이다.proof. In order to satisfy the second constraint of Problem 1 and [Equation 16] above, the computational capacity cannot have a negative value,

condition must be met. Otherwise, no feasible solution exists. That is, since the user's optimal strategy is not to offload, the optimal feasible solution is (0, 0).

가 실현 가능한 범위라면,

은 보조정리 6을 기준으로 하는 것이 최적이다.

is within the feasible range,

is optimally based on Lelem 6.

가 적으면 수정 비용 함수를 최소화한 해결책이 사용자의 실행지연 요건을 충족하지 못하는 경우다. 이 경우 사용자는 요건을 충족하기 위해 서버에서 최소한 F_min 계산 용량을 사용해야 한다. 비용 함수는 보조정리 5를 기반으로 한 볼록 함수이기 때문에 다음과 같은 불균등이 유지된다.Than F _min

If is small, the solution that minimizes the modified cost function does not meet the user's execution delay requirement. In this case, the user must use at least F _min computing capacity on the server to meet the requirement. Since the cost function is a convex function based on Lemma 5, the following inequality holds.

은 사용자 비용을 최소화하는 것이 실현 가능한 최적의 해결책이다. So in this case

is the optimal feasible solution to minimize user cost.

가 크면 서버의 남은 용량이 부족하거나 사용자가 오프로드로 혜택을 볼 수 없는 경우다. 같은 방법으로 다음과 같은 불균등이 지속된다.than F _max

If is large, either the server has insufficient remaining capacity or the user cannot benefit from offloading. In the same way, the following inequalities persist.

가 F_max보다 큰 경우

는 사용자 비용을 최소화하는 실현 가능한 최적 해결책이다.thus

is greater than F _max

is a feasible optimal solution that minimizes user cost.

는 서버가 사용자에게 가격 입찰 기능을 제안했을 때 최적의 오프로드 결정이다.

일 때 사용자는 비용을 최소화할 수 있으며, 이 결정에 근거한 오프로딩으로 실행 지연 요건을 충족할 수 있다.

일 때, 사용자는 오프로드 결정 체계에 기초하여 오프로드 태스크를 수행할 수 없다는 것을 알 수 있다. 따라서 이러한 오프로드 의사결정 방식에 기초한 오프로드 태스크는 사용자의 최적의 전략이다.this

is the optimal offload decision when the server offers the price bidding function to the user.

When , the user can minimize cost, and offloading based on this decision can meet the execution delay requirement.

When , the user may know that the offload task cannot be performed based on the offload decision scheme. Therefore, an offload task based on this offload decision-making method is an optimal strategy for the user.

B. Stackelberg equilibrium strategy

와

는 상술된 사용자의 최적 오프로드 결정이다. 서버가 a와 b를 변경하면

와

가 함께 변경된다. a와 b가 너무 크면 사용자는 약간의 오프로드가 발생하며, 실행 지연 요건을 충족하지 못할 수 있다. 그러면 사용자가 오프로드하지 않을 수 있다. 결과적으로, 서버의 유틸리티가 줄어들 수 있다.In this section, we propose a Stackelberg equilibrium strategy for servers. The purpose of the server is to maximize its utility as described above when the server knows the user's optimal offload strategy. Here, we assume that the user always makes offload decisions based on the optimal strategy, and the server knows the user's strategy. When the server proposes a price bid function, the user makes an optimal offload decision based on the proposed price bid function. Accordingly, the server must determine an appropriate price bidding function to maximize its utility. In this paper, the price bidding function is regarded as a linear function expressed as f(x) = ax+b, and the input of the price bidding function is the user's computational resource usage rate on the server. Therefore, the server must determine a and b, and these are the server's decision variables.

and

is the user's optimal offload decision described above. If the server changes a and b

and

are changed together. If a and b are too large, the user will experience some offload and may not meet the execution delay requirements. Then the user may not offload. As a result, the server's utility may be reduced.

In this regard, the server optimization problem to maximize its utility based on the user's optimal strategy can be expressed as follows.

Problem 3 (price bid function determination problem).

In various embodiments, supervised learning is proposed to find the function that determines the appropriate coefficients a and b. In practice, the server must quickly determine the price bidding function. By training the model to have the server find the appropriate coefficient based on server and user conditions, the server side can immediately suggest a price bidding function whenever the user wants to offload.

A feed-forward neural network is used for regression analysis based on supervised learning, and the configuration of the neural network is shown in FIG. 5 .

Different users have different local computing capacity, data size, and willingness to pay. Additionally, the number of CPU cycles required to calculate one bit of data and the execution latency requirements depend on the application used by each user. Therefore, users need these five types of information to make offload decisions. Since the state of server resources changes each time a user attempts to offload, the state of server resources must also be taken into account when determining the price bidding function. Therefore, the inputs of the model to find the appropriate coefficients a and b include these six elements. The outputs of the model must be the coefficients of the price bidding function, so a and b are the outputs of the model.

The data may have exceptional users, and these exceptional users may need to be offloaded unconditionally. Since the purpose of the server is to maximize revenue, this specially sensitive learning can result in less than optimal learning for average users. Therefore, an optimizer that is not sensitive to outliers should be used. The Adam optimizer is less sensitive to outliers because it uses past information as well as current information to learn. In some embodiments the Adam optimizer has a learning rate of 0.001 and the L1 loss is used for the loss function.

Generally, the server has information about offloading its history of coverage. That is, the server knows the distribution of inputs within coverage based on past information. The server can then generate data for training and perform an exhaustive search to find suitable a and b. In this task, 50,000 virtual users are created to create a data set, and an exhaustive search is conducted to create corresponding labels.

The overall process of determining the price bidding function when a server receives a new offload request after training is presented in an algorithm as shown in FIG. 6 . When a user wants to offload, the user sends an offload request including user and application information to the server. The server will receive that information and use it as input to a trained model of the server state. Finally, the trained price-bid function decision model returns the coefficients of the price-bid function.

7 is a diagram illustrating a computer system and an electronic device of a system model according to various embodiments.

Referring to FIG. 7 , the system model may include a computer system 710 and an electronic device 720 . The computer system 710 includes the aforementioned server, and the electronic device 720 refers to the aforementioned user and may include various types of IoT terminals.

The computer system 710 may include a communication module 711 , a memory 715 and a processor 717 . In some embodiments, at least one of the components of computer system 710 may be omitted and at least one other component may be added. In some embodiments, at least any two of the components of computer system 710 may be implemented as a single integrated circuit.

The communication module 711 may communicate with an external device in the computer system 710 . The communication module 711 may establish a communication channel between the computer system 710 and an external device and perform communication with the external device through the communication channel. The communication module 711 may include at least one of a wired communication module and a wireless communication module. The wired communication module may be connected to an external device through wired communication. The wireless communication module may include at least one of a short-distance communication module and a long-distance communication module. The short-distance communication module may communicate with an external device in a short-range communication method. The remote communication module may communicate with an external device through a remote communication method. Here, the remote communication module may communicate with an external device through a wireless network. In this case, the external device may include the electronic device 720 .

The memory 715 may store various data used by at least one component of the computer system 710 . For example, the memory 715 may include at least one of volatile memory and non-volatile memory. The data may include at least one program and related input data or output data. The program may be stored as software including at least one instruction in memory 715 .

The processor 717 may control at least one component of the computer system 710 by executing a program in the memory 715 . Through this, the processor 717 may perform data processing or calculation. At this time, the processor 717 may execute instructions stored in the memory 715. Also, the processor 717 may include a CPU. The CPU may process a task offloaded from the electronic device 720 .

The processor 717 may determine a price bidding function for setting a unit price per second for CPU usage of the computer system 710 for the electronic device 720 to be offloaded, and propose it to the electronic device 720. there is. Also, the processor 717 may receive CPU processing time of the computer system 710 for use in the electronic device 720 from the electronic device 720 . Through this, the processor 717 may be configured to process a task offloaded from the electronic device 720 through the CPU during CPU processing time.

The electronic device 720 may include a communication module 721 , an interface module 723 , a memory 725 and a processor 727 . In some embodiments, at least one of the components of the electronic device 720 may be omitted, and at least one other component may be added. In some embodiments, at least any two of the components of the electronic device 720 may be implemented as an integrated circuit.

The communication module 721 may perform communication with an external device in the electronic device 720 . The communication module 721 may establish a communication channel between the electronic device 720 and an external device and perform communication with the external device through the communication channel. The communication module 721 may include at least one of a wired communication module and a wireless communication module. The wired communication module may be connected to an external device through wired communication. The wireless communication module may include at least one of a short-distance communication module and a long-distance communication module. The short-distance communication module may communicate with an external device in a short-range communication method. The remote communication module may communicate with an external device through a remote communication method. Here, the remote communication module may communicate with an external device through a wireless network. At this time, the external device may include the computer system 710.

The interface module 723 may interface with a user of the electronic device 720 . The interface module 723 may include at least one of an input module and an output module. The input module may input a signal to be used in at least one component of the electronic device 720 . The input module includes an input device configured to allow a user to directly input a signal into the electronic device 720, a sensor device configured to sense a surrounding environment and generate a signal, or an image captured to generate image data. It may include at least one of the camera modules. The output module may include at least one of a display module for visually displaying information and an audio module for outputting information as an audio signal.

The memory 725 may store various data used by at least one component of the electronic device 720 . For example, the memory 725 may include at least one of volatile memory and non-volatile memory. The data may include at least one program and related input data or output data. The program may be stored as software including at least one instruction in memory 725 .

The processor 727 may control at least one component of the electronic device 720 by executing a program in the memory 725 . Through this, the processor 727 may perform data processing or calculation. At this time, the processor 727 may execute instructions stored in the memory 725. Also, the processor 727 may include a CPU.

The processor 727 may receive a price bid function for setting a unit price per second for CPU usage of the computer system 710 from the computer system 710 for use in the electronic device 720 . Also, the processor 727 may determine a CPU processing time of the computer system 710 to be used in the electronic device 720 based on the price bidding function, and transmit the determined CPU processing time to the computer system 710 . The CPU processing time may be determined based on the processor 727, the delay time condition for the application running on the electronic device, and the payment according to the CPU use of the computer system 710. Through this, the processor 727 may offload a task to the computer system 710 during CPU processing time.

According to various embodiments, a user of the electronic device may pay the computer system a payment that is determined as the product of a unit price per second and a CPU processing time.

8 is a diagram illustrating a method of a computer system 710 in accordance with various embodiments. 8 may represent an operation of a server as shown in FIG. 2 .

Referring to FIG. 8 , the computer system 710 may detect an electronic device 720 to be offloaded in step 810 . At this time, the processor 717 receives information about the electronic device 720 (US in FIG. 6) and information about an application running in the electronic device 720 (AS in FIG. 6) through the communication module 711. and the electronic device 720 can be detected based on this. Information about the electronic device 720 (US in FIG. 6 ) may include a local CPU capacity F _l of the electronic device 720 and an amount of data (d) to be processed by the electronic device. Information on the application (AS of FIG. 6 ) may include a latency condition (t _req ) for the application and a workload (h) condition for the application.

Computer system 710 may propose a price bid function for electronic device 720 in step 820 . The processor 717 calculates a price bid function based on information about the electronic device (US in FIG. 6), information about the application (AS in FIG. 6), and information about the computer system 710 (SS in FIG. 6). can decide Also, the processor 717 may transmit the price bidding function to the electronic device 720 through the communication module 711 .

The computer system 710 may receive the CPU processing time from the electronic device 710 in step 830 . The processor 717 may receive CPU processing time from the electronic device 710 through the communication module 711 . In response, the computer system 710 may allocate resources to the electronic device 720 in step 840 . The processor 717 may allocate resources to the electronic device 720 based on CPU processing time.

The computer system 710 may process the offloaded task from the electronic device 720 in step 850 . As the task is offloaded through resources from the electronic device 720 through the communication module 711 , the processor 717 may process the task of the electronic device 720 through the CPU. At this time, the processor 717 may process the task during CPU processing time. The computer system 710 may transmit the processing result to the electronic device 720 in step 860 . When the CPU processing time ends, the processor 717 may check the processing result for the task. And, the processor 717 may transmit the processing result to the electronic device 720 through the communication module 711 .

9 is a diagram illustrating a method of an electronic device 720 according to various embodiments.

Referring to FIG. 9 , the electronic device 720 may request an offload from the computer system 710 in step 910 . At this time, the processor 727 may check information about the electronic device 720 (US in FIG. 6 ) and information about an application running in the electronic device 720 (AS in FIG. 6 ). And, the processor 727 transmits information about the electronic device 720 (US in FIG. 6) and information about an application running in the electronic device 720 (AS in FIG. 6) through the communication module 711. While transmitting, an offload request may be made to the computer system 710 . Information about the electronic device 720 (US in FIG. 6 ) may include a local CPU capacity F _l of the electronic device 720 and an amount of data (d) to be processed by the electronic device. Information on the application (AS of FIG. 6 ) may include a latency condition (t _req ) for the application and a workload (h) condition for the application.

The electronic device 720 may receive a price bid function from the computer system 710 in step 920 . Processor 727 may receive a price bid function from computer system 710 via communication module 721 . The price bid function uses computer system 710 to obtain information about the electronic device (US in FIG. 6), information about the application (AS in FIG. 6), and information about computer system 710 (SS in FIG. 6). Based on this, it can be determined.

The electronic device 720 may transmit the CPU processing time to the computer system 710 in step 930 . The processor 727 may determine the CPU processing time based on the delay time condition for the application running in the electronic device 720 and the payment according to the CPU use of the computer system 710 . Also, the processor 727 may transmit the CPU processing time to the computer system 710 through the communication module 721 .

The electronic device 720 may detect resources allocated from the computer system 710 in step 940 . The processor 727 may detect resources allocated from the computer system 710 through the communication module 721 . At this time, the computer system 710 may allocate resources to the electronic device 720 based on the CPU processing time.

The electronic device 720 may offload the task to the computer system 710 in step 950 . Processor 727 may offload tasks to computer system 710 via communication module 721 . At this time, the processor 727 may offload the task during CPU processing time.

The electronic device 720 may receive a processing result from the computer system 710 in step 960 . After the CPU processing time ends, the processor 727 may check the processing result of the task from the computer system 710 through the communication module 721 .

A method of a computer system according to various embodiments includes detecting an electronic device to be offloaded, proposing a price bidding function to set a unit price per second for CPU usage of the computer system for the electronic device, and including the electronic device. Receiving CPU processing time of the computer system for use in the electronic device from the computer system, and processing tasks offloaded from the electronic device during the CPU processing time.

According to various embodiments, the detecting of the electronic device may detect the electronic device based on reception of information about the electronic device and information about an application running in the electronic device.

According to various embodiments, the information on the electronic device may include a local CPU capacity of the electronic device and an amount of data to be processed by the electronic device.

According to various embodiments, the information about the application may include a latency condition of the application and a workload condition of the application.

According to various embodiments, the step of proposing a price bid function may determine a price bid function based on information about an electronic device, information about an application, and information about a computer system.

According to various embodiments, the electronic device may determine the CPU processing time based on a delay time condition for an application running in the electronic device and payment according to CPU use of a computer system.

According to various embodiments, a user of the electronic device may pay the computer system a payment that is determined as the product of a unit price per second and a CPU processing time.

According to various embodiments, processing the task may include allocating resources to the electronic device based on CPU processing time, and processing the task as the task is offloaded through the resources. can

A computer system according to various embodiments may include a memory and a processor connected to the memory and configured to execute at least one instruction stored in the memory.

According to various embodiments, the processor detects the electronic device it wishes to offload, proposes a price bid function to set a unit price per second for CPU usage of the computer system for the electronic device, and from the electronic device to the electronic device. It may be configured to receive CPU processing time of the computer system for use and process tasks offloaded from the electronic device during the CPU processing time.

According to various embodiments, the processor may be configured to detect the electronic device based on receiving information about the electronic device and information about an application running in the electronic device.

According to various embodiments, the information on the electronic device may include a local CPU capacity of the electronic device and an amount of data to be processed by the electronic device.

According to various embodiments, the information about the application may include a latency condition of the application and a workload condition of the application.

According to various embodiments, the processor may be configured to determine a price bidding function based on information about the electronic device, information about the application, and information about the computer system.

According to various embodiments, the electronic device may determine the CPU processing time based on a delay time condition for an application running in the electronic device and payment according to CPU use of a computer system.

According to various embodiments, a user of the electronic device may pay the computer system a payment that is determined as the product of a unit price per second and a CPU processing time.

According to various embodiments, the processor may be configured to allocate resources to the electronic device based on CPU processing time and process the task as the task is offloaded through the resources.

The devices described above may be implemented as hardware components, software components, and/or a combination of hardware components and software components. For example, devices and components described in the embodiments include a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), and a programmable PLU (programmable logic unit). logic unit), microprocessor, or any other device capable of executing and responding to instructions. The processing device may run an operating system (OS) and one or more software applications running on the operating system. A processing device may also access, store, manipulate, process, and generate data in response to execution of software. For convenience of understanding, there are cases in which one processing device is used, but those skilled in the art will understand that the processing device includes a plurality of processing elements and/or a plurality of types of processing elements. It can be seen that it can include. For example, a processing device may include a plurality of processors or a processor and a controller. Other processing configurations are also possible, such as parallel processors.

Software may include a computer program, code, instructions, or a combination of one or more of the foregoing, which configures a processing device to operate as desired or processes independently or collectively. You can command the device. The software and/or data may be embodied in any tangible machine, component, physical device, computer storage medium or device to be interpreted by or to provide instructions or data to a processing device. there is. Software may be distributed on networked computer systems and stored or executed in a distributed manner. Software and data may be stored on one or more computer readable media.

Methods according to various embodiments may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer readable medium. In this case, the medium may continuously store a program executable by a computer or temporarily store the program for execution or download. Also, the medium may be a single or various types of recording means or storage means in the form of a combination of several pieces of hardware. It is not limited to a medium directly connected to a certain computer system, and may be distributed on a network. Examples of the medium include magnetic media such as hard disks, floppy disks and magnetic tapes, optical recording media such as CD-ROM and DVD, magneto-optical media such as floptical disks, and ROM, RAM, flash memory, etc. configured to store program instructions. In addition, examples of other media include recording media or storage media managed by an app store that distributes applications, a site that supplies or distributes various other software, and a server.

Various embodiments of this document and terms used therein are not intended to limit the technology described in this document to a specific embodiment, and should be understood to include various modifications, equivalents, and/or substitutes of the embodiment. In connection with the description of the drawings, like reference numerals may be used for like elements. Singular expressions may include plural expressions unless the context clearly dictates otherwise. In this document, expressions such as "A or B", "at least one of A and/or B", "A, B or C" or "at least one of A, B and/or C" refer to all of the items listed together. Possible combinations may be included. Expressions such as "first," "second," "first," or "second" may modify the elements in any order or importance, and are used only to distinguish one element from another. The components are not limited. When a (e.g., first) element is referred to as being "(functionally or communicatively) connected" or "connected" to another (e.g., second) element, it is referred to as being "connected" to the other (e.g., second) element. It may be directly connected to the component or connected through another component (eg, a third component).

The term "module" used in this document includes a unit composed of hardware, software, or firmware, and may be used interchangeably with terms such as logic, logic block, component, or circuit, for example. A module may be an integral part or a minimum unit or part thereof that performs one or more functions. For example, the module may be composed of an application-specific integrated circuit (ASIC).

According to various embodiments, each component (eg, module or program) of the described components may include a singular object or a plurality of entities. According to various embodiments, one or more components or steps among the aforementioned components may be omitted, or one or more other components or steps may be added. Alternatively or additionally, a plurality of components (eg modules or programs) may be integrated into a single component. In this case, the integrated component may perform one or more functions of each of the plurality of components identically or similarly to those performed by the corresponding component among the plurality of components prior to integration. According to various embodiments, steps performed by a module, program, or other component are executed sequentially, in parallel, iteratively, or heuristically, or one or more of the steps are executed in a different order, omitted, or , or one or more other steps may be added.

Claims (15)

In the method of the computer system,
detecting an electronic device to be offloaded;
proposing a price bidding function to set a unit price per second for CPU usage of the computer system for the electronic device;
receiving, from the electronic device, CPU processing time of the computer system for use in the electronic device; and
Processing a task offloaded from the electronic device during the CPU processing time.
including,
Detecting the electronic device,
Detecting the electronic device based on reception of information about the electronic device and information about an application running in the electronic device;
The information on the electronic device includes a local CPU capacity of the electronic device and an amount of data to be processed by the electronic device;
The information on the application includes a latency condition for the application and a workload condition for the application;
The step of proposing the price bidding function,
determine the price bidding function based on the information about the electronic device, the information about the application, and the information about the computer system;
The electronic device,
determining the CPU processing time based on a latency condition for the application and a payout according to CPU usage of the computer system;
method.
delete delete delete delete According to claim 1,
The user of the electronic device,
Paying to the computer system a payment determined by the product of the unit price per second and the CPU processing time.
method.
According to claim 1,
The step of processing the task is,
allocating resources to the electronic device based on the CPU processing time; and
processing the task as the task is offloaded through the resources;
including,
method.
In a computer system,
Memory; and
a processor coupled with the memory and configured to execute at least one instruction stored in the memory;
the processor,
detect an electronic device to be offloaded;
Propose a price bidding function for setting a unit price per second for CPU usage of the computer system for the electronic device;
Receive from the electronic device CPU processing time of the computer system for use in the electronic device;
During the CPU processing time, configured to process a task offloaded from the electronic device,
the processor,
configured to detect the electronic device based on reception of information about the electronic device and information about an application running in the electronic device;
The information on the electronic device includes a local CPU capacity of the electronic device and an amount of data to be processed by the electronic device,
The information on the application includes a latency condition for the application and a workload condition for the application;
the processor,
determine the price bid function based on the information about the electronic device, the information about the application, and the information about the computer system;
The electronic device,
determining the CPU processing time based on a latency condition for the application and a payout according to the CPU usage of the computer system;
computer system.
delete delete delete delete According to claim 8,
The user of the electronic device,
Paying to the computer system a payment determined by the product of the unit price per second and the CPU processing time.
computer system.
According to claim 8,
the processor,
Allocating resources to the electronic device based on the CPU processing time;
configured to process the task as it is offloaded through the resources;
computer system.
A non-transitory computer-readable storage medium comprising:
detecting an electronic device to be offloaded;
proposing a price bidding function to set a unit price per second for CPU usage of a computer system for the electronic device;
receiving, from the electronic device, CPU processing time of the computer system for use in the electronic device; and
Processing a task offloaded from the electronic device during the CPU processing time.
Store one or more programs for executing,
Detecting the electronic device,
Detecting the electronic device based on reception of information about the electronic device and information about an application running in the electronic device;
The information on the electronic device includes a local CPU capacity of the electronic device and an amount of data to be processed by the electronic device;
The information on the application includes a latency condition for the application and a workload condition for the application;
The step of proposing the price bidding function,
determine the price bidding function based on the information about the electronic device, the information about the application, and the information about the computer system;
The electronic device,
determining the CPU processing time based on a latency condition for the application and a payout according to CPU usage of the computer system;
A computer-readable storage medium.

Images